CAREER: Capturing the real world for computer graphics

数码相机和光学技术的进步正在迅速改变计算机图形学和计算机辅助几何设计的面貌。就在几年前，这些领域主要是由艰苦的用户交互驱动的。物体及其表面特性几乎完全由手工建模，或者使用扫描的颜色纹理、脆弱的计算机视觉技术和基于触摸探针的慢速形状测量方法从真实测量中重建。在过去的十年里，主要是由于几何稳定的成像技术的可用性，如电荷耦合器件（CCD），紧凑的光源，如激光二极管，以及具有成本效益的定制光学器件，将现实世界带入计算机的能力以爆炸式的速度增长。利用这些技术，我们可以获得大大小小的物体的形状和外观。这种能力在许多领域都有应用，包括机械设计，逆向工程，工业检验，医学，特效和游戏，以及博物馆文物的传播。这个项目将研究一些将现实世界的形状和外观带入计算机的方法。在形状捕获领域，研究将跨越3D扫描管道：从改进的测距方法，到更快、更简洁的配准和重建算法。此外，传统的2D图像将用于在各个阶段增强管道。在外观捕捉领域，将探索两个新的方向：在表面上建模和渲染亮度函数，以及基于样本的方法，用于表示未知几何形状物体反射和折射光线的方式。第一个方向有望克服当前纯基于图像的渲染方法的一些限制，而后者代表了在真实物体和合成物体之间建模光传输的新范式。该项目的教育目标包括让毕业生，特别是本科生参与教学和研究。此外，还将开设新的“3D摄影”课程。通过讲座和实践项目，学生将学习捕捉现实世界物体的形状和外观的原理和实践，然后将它们转换成适合工业设计和计算机动画的表示形式。

Advances in digital camera and optical technology are rapidly changing the face of computer graphics and computer-aided geometric design. Just a few years ago, these domains were primarily driven by painstaking user interaction. Objects and their surface properties were modeled almost exclusively by hand or were reconstructed from real measurements using scanned color textures, fragile computer vision techniques, and slow shape measurement methods based on touch probes. In the past decade, largely due to the availability of geometrically stable imaging technology such as charge-coupled devices (CCD's), compact light sources such as laser diodes, and cost-effective custom optics, the ability to bring the real world into the computer has grown at an explosive rate. Using these technologies we can acquire the shape and appearance of objects large and small. This capability has applications in numerous fields includingmechanical design, reverse engineering, industrial inspection, medicine, special effects and games, and dissemination of museum artifacts.This project will investigate a number of methods for bringing the shape and appearance of the real world into the computer. In the area of shape capture, the research will span the 3D scanning pipeline: from improved range finding methods, to faster and more concise registration and reconstruction algorithms. Further, traditional 2D images will be used to augment the pipeline at all stages. In the area of appearance capture, two novel directions will be explored: modeling and rendering radiance functions over surfaces and sample-based methods for representing the manner in which objects of unknown geometry reflect and refract light. The first direction promises to overcome some of the limitations of current, purely image-based rendering methods, while the latter represents a new paradigm for modeling light transport among real and synthetic objects. The educational goals of the project include involving graduates and particularly undergraduates in both teaching and research. In addition, a new "3D photography" course will be developed. Through lectures and hands-on projects, students will learn the principles and practice of capturing the shape and appearance of real world objects and then converting them into representations suitable for industrial design and computer animation.